Re: orbit-idl-2 new option request, please comment ...

[Jules Colding <colding omesc com>]

Hi KC,

On Thu, 2005-08-11 at 05:15 +0800, KC wrote:
> I think it's better to have a new orbit-idl-2 option "--skeleton-impl-epv-only"
> which will only emit epv structure and nothing else.
> 
> I believe this is important and worth to have an special option because
> the methods' "order" in epv structure is very important for ORBit2 specially
> for ORBit_small_invoke_async() ... since we need the index
> of of method for Foo__iinterface.methods._buffer[] and there are NO
> support for async stuff in orbit-idl-2 yet.
> 
> Other part of "--skeleton-impl" output is not so critical ..., well,
> I feel OK even coding them by hand. 
> 
> And some code emitted by "--skeleton-impl", IMHO, should not even be
> emitted ... such as /*** App-specific servant structures ***/ and
> /*** Stub implementation ***/
> I think emitting
> 
>     #include "app-servant.c.inc"
> and
>     #include "app-methods.c.inc"
> 
> make more sense .... since when user re-define IDL, he/she may re-invoke
> orbit-idl-2 again, and the code previously added will be overwritten.

The way I do it is to take the generated "foo-skelimpl.c" file and
rename it to "foo_impl.c". Whenever the interface changes I just copy
any changes from the new "skelimpl" file to the "_impl" one.

It is, IMHO, practical to have it all included in one file. Emitting
more files would be adding more administrative overhead to the project.
I have one exception to that though. See my answer to your other mail.

Best regards,
  jules